Getting latest samples from a running task

[mirode]

Hi all,

I've been struggling with the following for several days now, but cannot find a satisfactory solution so far.

I need to run an AI-task (nothing critical regarding speed, lets say c.a. 1kHz sample rate), that continuously reads channels on an AI card (I am using a 9201 in cDAQ 9188, but have tried an USB daq as well and the problems are just the same). Then, at some random times I need to get the last n samples from the buffer, while the acquisition continues. As this operation should be executed from different threads, it is possible that I execute overlapping readouts as well.

I need all this to be able to get the data quickly, from different areas of my program, but still enable some integration over the samples.

I am using the DAQmx read properties of "Relative to" (last sample) and "Offset" (-n), followed by a standard DAQmx read analog VI, but what I am getting is samples which are too old, values that remain unchanged for multiple readouts or just errors, depending how I configure the timing of the task.

It is just a simple requirement - read last n values from a continuous task, but I cannot make it work.

Thank you in advance for your suggestions!

Cheers,

mirode

PS. LV 2012, DAQmx 9.6.1f1

[crossrulz]

I would recommend having a separate thread to handle all of the reading from the DAQmx channel.  Have it read the data and save it in a circular buffer.  I would actually recommend using a Functional Global Variable (aka LV2 Globals, Action Engines).  Your other threads can read this data from the FGV whenever they need the data.

[mirode]

Hi crossrulz,

thanks for your fast response. In fact I was thinking of such a solution, but cannot believe there is no way to utilize the DAQmx buffer which is just there...

[crossrulz]

Mucking with the DAQmx buffer and reading it from multiple threads scares me.  I can easily see why you could get stale data.  Besides, you really should only have 1 place reading from hardware.  It makes things a lot cleaner and scalable.

[mirode]

I understand your point, but still I AM using 1 point buffer access, with a single sub-VI which is not reentrant.

Anyway I begin to like your idea with the circullar buffer.

[Mark_Yedinak]

Another solution would be to post your latest data to a notifer. Your read task can post the data to the notifer and the readers will always get the most recent data from when the process the notification.

[mirode]

Thanks guys, I am trying out both methods to see which one best integrates with the rest of my software and the tests so far are very promising.